Chat UI

Expand

The Chat page (/chat) is the primary way to interact with Snippbot agents. It supports streaming responses, file uploads, multi-agent selection, model switching, and inline project approval.

Interface layout

The chat page uses a resizable multi-pane layout:

• Left pane — Conversation history sidebar and workspace file tree
• Center pane — Main chat interface with message stream, input bar, and attached files
• Right pane — Tool timeline showing real-time tool activity during agent execution

The panes are resizable by dragging the dividers, and the right pane (tool timeline) can be collapsed.

Agent selector

Click the agent selector at the top to switch which agent handles your messages. Each agent has its own persona, model, and memory.

Model selector

Next to the agent selector, a model picker lets you override the agent’s default model for the current conversation. Available providers include Anthropic, OpenAI, OpenRouter, Gemini, DeepSeek, xAI Grok, Groq, and Mistral.

Sending messages

Type a message and press Enter or click Send. Use Shift+Enter to insert a new line. The response streams token-by-token using SSE.

To describe a multi-step goal, be specific:

Research the top 5 Python web frameworks by GitHub stars.
For each one, note the license, primary use case, and latest release date.
Output a markdown table.

File upload

Click the paperclip icon or drag files into the drop zone to attach files. Supported types:

• Text/code: .txt, .py, .js, .ts, .json, .yaml, .toml, .md
• Documents: .pdf
• Images: .png, .jpg, .gif, .webp (for vision-capable models)

Attached files are sent as context with your message. An asset picker modal also allows selecting from previously uploaded assets.

Image mode selector

For vision-capable models, an image mode selector lets you control how attached images are processed — useful for choosing between detail levels.

Project plans

When a goal requires multiple steps, the agent generates a project plan before executing. The plan appears inline with:

• Phase and task breakdown
• Estimated steps per phase
• Approve button to start execution
• Edit to modify tasks before approving

A Generate Project modal also allows converting a conversation into a structured project with phases and tasks.

Tip

Review the plan before approving. You can cancel individual tasks, reorder phases, or edit task descriptions before execution begins.

Conversation history

Click the history dropdown (top area) to see past conversations. Each conversation is stored and searchable. The workspace history panel shows files created or modified during conversations.

Mode toggle

The Mode Toggle switches between conversation modes (e.g., Plan mode vs. Execute mode), affecting how the agent processes your requests. A Plan Mode Banner appears when plan mode is active.

Stopping a response

Click Stop (appears during streaming) to interrupt the current response. The partial output is saved.

Session management

A Session Indicator shows the current session state. For remote sessions, you can transfer control to another device using the Transfer Modal and approve sessions via the Session Approval Modal.